Automated detection of cylindrical structures in complex pipelines using iterative point cloud segmentation and...

Area of Science:

• Computer Vision
• Geometric Modeling
• Reverse Engineering

Background:

• Cylinders are essential structural components in pipeline systems.
• Accurate detection of cylinders from 3D point clouds is vital for efficient reverse engineering.
• Current methods face limitations with complex pipeline geometries.

Purpose of the Study:

• To develop a novel, automated method for robust cylinder detection in unstructured 3D point clouds.
• To overcome the limitations of existing methods in handling complex pipeline scenarios.
• To enhance the automation of reverse engineering for intricate pipeline designs.

Main Methods:

• Iterative clustering segmentation for data complexity reduction.
• Three-point random sampling for reliable candidate cylinder estimation.
• High-precision cylinder fitting and multi-filtering for minimizing false positives.

Main Results:

• The proposed method demonstrates superior performance in cylinder detection.
• Achieved precision of 0.8727, recall of 0.8090, and F1 score of 0.8397.
• Outperformed existing cylinder detection techniques on both simulated and real-world datasets.

Conclusions:

• The novel method offers an accurate and robust solution for automated cylinder detection.
• This approach significantly advances the reverse engineering process for complex pipelines.
• The findings highlight the method's potential for practical industrial applications.